Very long calculation. Pipeline issue?

[Sebu]

I notice that the calculation time is very, (very, very) long.
And yet I have a decent machine configuration.
So I'm wondering if there isn't a bottleneck somewhere in my config?
Here https://filebin.net/bogs5suy1vwgnab8 is the log of a current calculation of 580 cameras started 2 days ago.
Still calculating.
Why do all the computer components run at a minimum in this final phase?
CUDA activated.
Do you have any idea? Or maybe the calculation time is normal.
Curious to know what the problem is!
Thanks for your advice,
Sebu.

[Sebu]

Well, after 3 days, I finally cancelled the mesh operation on the dense cloud.

https://filebin.net/cdt53orh8pkhwnf7

I'm going to try to divide the mesh into batches of photos.

GPU is up to date, with CUDA enabled.

https://benchmark.3dflow.net/photogrammetry-benchmark

My config :
i7-13700K
RTX 3080 Ti
RAM 65300 Mo
Disk SSD

I think the weak link is the CPU.
Does anyone have any experience in this area?
Sebu

[Andrea Alessi]

Hi Sebu,

these processing times are very strange, it doesn't make sense that you have that many hours for coloring. The fact that it takes hours to do something like that makes me thing you're possibly having an I/O issue - where are these images located? Make sure they are on a local fast disk, I don't see anything particularly wrong with your settings or device, so I think this is something in your environment, either I/O or something interfering with zephyr.

[Sebu]

Thanks for your reply. I've done some tests and it turns out that my SSD drive isn't performing at all.
So I migrated the photos and the unified and persistent cache to the high-performance NVMe C: drive and, oh miracle, it works like a charm!